Data Center Execution Assurance
Building on the trust assumption of the attestation flow, we want to investigate how to provide a proof of cloud. This ensures that when interacting with a TEE, we can also obtain assurance that it is operated from a cloud data center. One approach consists of relying exclusively on cryptography to anchor the said TEE to the hardware roots of trusts
of machines known to belong and be operated by trusted cloud providers. Another approach consists of obtaining a rough location estimate (cloud providers' data centers are usually known) from protocols such as GoAT [3], and therefore basing strong trust in a given geographical location.
In a second phase, the project could leverage these new location guarantees to enable a geographically decentralized global MPC network. This network would incentivize geographically dispersed nodes by increasing rewards, such as a fraction of the MEV in the block, for underrepresented regions. Drawing inspiration from decentralized file storage systems like GoAT [3], the project will explore proof-of-geo-retrievability-like methods to verify the geographic location of participating nodes, enhancing overall network decentralization.
Problem Description
- Understanding of the different attestation models of TEEs for local/bare metal and cloud deployments
- and providing a “Proof of Cloud” that ensures we communicate with a TEE running in a cloud.
Bare metal deployments and cloud deployments provide different trust guarantees when it comes to attestation flows. Therefore, we aim to investigate the attestation models of Intel TDX and AMD SEV, as well as cloud-provided attestations such as Microsoft Azure Attestation (MAA) and Google Cloud's vTPM [2]. The research will investigate how these attestations can be combined with native TEE attestations to build a more comprehensive trust model [1]. Techniques like zkTLS may be explored to ensure the correctness of queries to attestation services without additional reliance on TEEs. The main goal is to argue whether they provide the same guarantees on the attestation flow.
References
Vision paper
Presentation at the SysTEX workshop co-located with EuroS&P on 4 July.
Link to the paper: https://arxiv.org/pdf/2506.14964
The vision paper describes the problem the Proof-of-Cloud aims to solve, and integrates the previous work of the Andromeda team towards a proof of cloud. It relies on a PPID provided by the cloud provider. The cloud provider keeps a track of active VMs in their cloud and provides this list as a part of the attestation flow. Ideally, all cloud providers should use the same PPID protocol to ensure compatibility among them. The challenge is the need to trust the cloud providers that they provide a correct list.
